[codex] Add ELF-driven model fixtures and trace checks

[zhoubot]

What changed

• add the ELF/program-image loader, Minst decode/disassembly stack, and the reference/compare CLI path in tools/model
• add richer Minst trace decoration and objdump-style .dump output next to JSONL traces
• add committed LLVM-built ELF fixtures plus a manifest/regeneration script, and a ctest check that runs linx_model_cli against those ELF inputs

Why

• the model tests previously relied on handcrafted binaries instead of real LLVM-produced ELF inputs
• we need deterministic, repo-owned fixture coverage that exercises the ELF loader, disassembler, and reference executor on the same examples every run
• this also closes the loop for the current AVS model smoke sources and keeps the trace/schema path under test

Impact

• ctest now includes linx_model_check_llvm_elf_fixtures
• fixture regeneration is documented and scriptable from a superproject checkout
• Minst trace output carries canonical JSON fields plus objdump-style dump lines for review/debugging

Validation

• ctest --test-dir /Users/zhoubot/linx-isa/tools/model/build --output-on-failure
• python3 /Users/zhoubot/linx-isa/tools/model/tests/fixtures/llvm_elf/regenerate_llvm_elf_fixtures.py --root /Users/zhoubot/linx-isa/tools/model

[gemini-code-assist]

Warning

Gemini is experiencing higher than usual traffic and was unable to create the review. Please try again in a few hours by commenting /gemini review.